There’s an interesting article in The Guardian comparing compressiom rates and qualities for voice-based audio files. This is useful when considering producing audio for interactive environments and devices, as well as content that is shared via online audio services.
It’s an interesting subject and one which has changed a lot over the years. We deal with just 2.4kbits/s with two way radio but yet 500kbits/s for broadcast links. Perhaps as speaker quality improves and data speeds and storage capacities get bigger then these low bitrates will be irrelevant.
I know we do some streaming for speech only services at 16kbit/s and nobody complains about the quality.